Think I have a dud card or insufficient PSU. Want to run it by the forums. I have a GeForce RTX 3080 Ti VISION OC 12G which is under warranty. When playing certain games where the card would be under full load the computer will reboot randomly. This was the case for both of the machines listed below. No OC, no funny stuff.
Tested in updated W10 and Ubuntu MATE. Most recent drivers for W10 and nvidia-driver-510 on MATE (the newer ones caused a weird screen flicker but that’s a video for another time.) Used DDU on W10, reinstalled drivers, etc.
Now the card DOES work when I enable active State Power Management function (ASPM) on the Aorus Master. At this rate I was just praying something worked and this appeared to.
The thing is, I have no idea what it is. Manual is light on detail… No idea if it is limiting power draw, something that should be on by default, if my card should work without it, etc etc.
Let me know your thoughts on if its GPU/PSU, whatever the hell ASPM is, and if I should bother with Gigabyte support. I have heard horror stories. Engagement challenge, tell me your experience.
I had random reboots and crashes with my 3080Ti on a (what I think is) a perfectly good 850W PSU. Pretty much everything about my system is different from yours, apart for being a 3080Ti. Nothing in the Windows system logs, nothing in the UEFI helped. Then I saw the Gamers Nexus vid about power spikes on high-end Nvidia cards.
Replaced the PSU with a 1000W version and the problem totally went away.
Bottom line: You might need to use more than a 850W PSU with a 3080Ti under heavy load.
I was talking to my friend when that video came out and how nVidia and the power supply companies are both using the bare minimum of capacitance and blaming the other. I mentioned how I had seen people DIY capacitors in right at the power connector and there would probably be an aftermarket for some ungodly RGB power conditioner and he showed me that they already exist:
They don’t have one specifically for your card, but I’m thinking this is going to be the future for graphics cards. Buying an absurdly oversized PSU for those odd fractions of a second of increased current draw is going to hit a wall when you can only draw a maximum of 1800 watts from a 15 amp branch circuit at 120 volts.
Want to follow up here after a long time of troubleshooting in case it helps anyone prior to spending more money to potentially fix a problem.
Windows implementation is something to be desired when setting power limits on graphics cards. They don’t seem to be honored, or at least in my case. Tried via 3 different methods. It may also be me.
Linux is a little bit of a different story
via nvidia-smi I was able to knock the power limit of the 350W base power of the card to 200W and then incrementally 250W, stable.
From the command line did so with
sudo nvidia-smi -pl 200
If I remember correctly monitored power consumption via
nvidia-smi -pm 1
nvidia-smi stats -i pwrDraw
or some variation thereof.
Conclusion, power supply. Thanks Nvidia for the power draw. 1300W Super Flower psu for $180 on the way. Overkill on purpose, have some other things in the mix.
Last to note, this might be game by game specific. When reducing the Power Limit to play the Witcher 3, I saw minimal change in performance. I may be speaking out of turn as an idiot but I wonder if the game isn’t the most optimized. where the heck is the next-gen release cdprojekt red
Hope this helps someone as the internet is riddled with information. Hopefully this works as a quick check before going own the rabbit hole.